Multi-object tracking is an important part of computer vision and plays a crucial role in many real-world application scenarios. The existing tracking frameworks have the problem of mismatching when dealing with highly similar objects in highly dense regions. In order to solve the above problems, this paper proposes a multi-target trajectory association algorithm based on appearance feature similarity clustering module and dynamic partition cascade matching. The algorithm uses the appearance similarity clustering module based on the customized DBScan algorithm. By calculating the Euclidean distance of the position coordinates and the cosine distance of the appearance features between the targets, the highly similar objects in the highly dense area are clustered. Then, the dynamic partition cascade matching strategy is used. Firstly, the dynamic partition numerical algorithm is used to calculate the partition value M × N of different objects according to the average similarity of features. Partition cascade matching is used to divide the “clustered” target into specific partitions, and the cosine distance of the feature vector is calculated with the trajectory to achieve a more refined matching. Comprehensive experiments are carried out on DanceTrack and MOT20 datasets, and the MOTA score of 92.3% is achieved on DanceTrack dataset, which shows that the proposed method has good tracking performance.